Toward high-precision analysis of soil micro-and nanoplastics: A review of spectroscopy and machine learning approaches

Abstract

Micro-nanoplastics (MNPs) in soil have been recognised as a threat to soil and environment, impacting soil and food security, human health and ecosystem services. Current methods for analysing MNPs in soil are time consuming and expensive and hence cannot meet the growing need to quantify this threat on a larger scale. This review summarises existing knowledge on MNPs in soil and current analytical approaches. This includes examining the opportunities and challenges associated with various spectroscopy methods, such as visible–near-infrared (Vis-NIR), mid-infrared (MIR), Raman, terahertz (THz), and hyperspectral imaging (HSI), in detecting and quantifying MNPs in soil through machine learning enhanced analysis. While these methods show promise as cost-effective methods to facilitate routine analysis, key challenges persist. These include a lack of validation on naturally contaminated samples, difficulty detecting nanoplastics, a focus on classification rather than quantification, and detection limits that often exceed environmentally relevant concentrations. The review recommends expanding spectral libraries, enhancing resolution through spectral fusion and HSI, and integrating soil properties into machine learning models to improve detection accuracy.